Skip to content
gitogi

AI integration

Post-service ratings: the views of those who leave them

Post-visit ratings can leave unhappy customers out of the picture. How to read repeat calls, job reports and renewals in your service business.

8 min read

This article was generated with AI assistance and may contain inaccuracies — verify critical information. How we use AI.

Post-service ratings: the views of those who leave them

In a maintenance business with around a dozen technicians, the day almost always ends the same way. The technician completes the visit, the customer signs the job report and, if they want to, leaves a rating. At the end of the month, the owner reviews the scores: they look good. Keeping track of contract expiry dates comes down to the memory of the people in the office who know the customers.

Then, in October, a contract isn't renewed. The customer had never complained, and the ratings collected suggested everything was fine. No one in the office can explain what happened.

This is a typical scenario, rather than the story of a specific business. If you run a service business, though, you probably recognise it. To understand what ratings can miss, it helps to start somewhere far removed from maintenance.

Businesses and customers see things differently

On 27 October 2026, Twilio, a company that provides customer service systems, discussed the findings of its own survey on its blog. It had surveyed 7,652 consumers across 18 countries, including Italy, who had bought something online in the previous six months. It had also surveyed 660 executives at companies with at least 500 employees. Responses were collected between 9 April and 1 May 2026.

The subject was AI-assisted customer service. While 86% of businesses believe their customers are satisfied, 51% of consumers agree. That leaves a gap of 35 percentage points between how service providers see the experience and how customers see it.

The blog links part of this gap to the way ratings are collected: a survey sent after an interaction only reaches the customers who are still there. Those who give up halfway through or stop replying may also be missing from the final ratings. There is another problem with how conversations are counted: an automated system counts a conversation as “handled” whether the issue was resolved or the customer simply stopped replying. If you only look at that total, the two outcomes appear identical.

The survey covers AI-assisted customer service, consumers who shop online and large companies. It does not measure satisfaction after a maintenance visit, but it raises a useful question if you run a service business: what do you know about the customers who stop replying?

When the customers who don't rate you are unhappy

Three researchers set out to understand what ratings were missing by analysing more than 170,000 customer service chat conversations at Samsung. At the end of each chat, customers could leave a rating of one to five stars.

Around one customer in six responded: 16.2%. Among those who left a rating, 68.1% gave four or five stars. Based on those scores alone, the verdict on the service looked positive. The remaining 83.8% had left the chat without giving a rating.

To estimate how these customers would have rated the service, the researchers used a program that had learned to associate the content of conversations with the scores given by customers who had responded. The estimate suggested that more than half of the silent customers, 54.5%, would have expressed dissatisfaction. The authors point to behaviour you will recognise if you run a business: many unhappy customers leave without complaining.

The study dates from 2018 and covers online customer service at a single company. Three human reviewers manually reviewed around a hundred chats, but the predictions cannot be fully verified because there are no actual ratings to compare them against. The result should therefore be read as an estimate, rather than an opinion those customers actually expressed. Silence does not amount to a positive rating, even when the ratings you collect are good.

What this means for a maintenance business

Think back to the business owner at the start. A customer who has had to call twice about the same fault may also leave no rating. They may have lost confidence and have no desire to explain why on the job report, preferring to wait until the contract expires. The first visible sign for the owner may then be the failure to renew, months later.

Consultancy Bain & Company had observed a similar pattern as far back as 2014, in an article about telecoms providers: many customers who left had been classified as satisfied. The article proposes a practical response. If a customer calls several times about a fault in the same week, a team of experts calls them back and can send someone out straight away. Repeat calls become a reason to act, without waiting for a rating.

The effect can also work in the other direction. mTab, a company that analyses customer surveys, explains that responses often come from very happy customers who want to say thank you and very angry customers who want a solution. Those in the middle tend to ignore the invitation. A survey sent after a support request has been closed can therefore make dissatisfaction look more widespread than it is. How and when you ask for a rating affects the responses: to decide what to do, you also need to know how your customers behave.

The signals you already have, and how to read them

To find what ratings leave out, you can start with the information you already collect. The blog we began with suggests measuring where customers drop out at each stage of the service. What follows is our way of applying that approach to maintenance work, from the request for a visit through to contract renewal.

In a maintenance business, your starting points are job reports, invoices and contracts. There are three signals to look for.

The first is open requests that never led to an appointment. A customer called, but nothing came of the request: are they still waiting, or have they already found help elsewhere?

The second is repeat visits for the same fault within a short period, for example within thirty days. If the customer called again because the problem returned, their rating of the first visit may not tell you what happened afterwards.

The third is contracts approaching expiry with no renewal offer. If one person keeps the dates in their head, even a week's holiday can mean missing the right moment to discuss renewal with the customer.

You can then check whether silence is associated with a difference in renewals. Take contracts eligible for renewal that expired during the same period and whose renewal outcomes are already observable. Include only customers who had the opportunity to leave a rating before their contracts expired, and separate those who rated by that date from those who did not. For each group, divide the number of renewed contracts by the total number of contracts eligible for renewal in that group. If the proportions are similar, this comparison shows no difference between the two groups. If customers who leave no rating renew much less often, you have a signal worth investigating: you can go back to job reports and repeat calls to try to understand what happened.

We do not have Italian data on response rates for post-visit ratings. This comparison, however, uses your own business data and helps you identify which customers to focus on.

Where we start

There is a practical obstacle to doing this work: the information may be available, but held separately. Service visits are in one place, contracts in another and invoices somewhere else. For the customer, though, they are all part of the same service. On our website, we explain it this way: contracts, equipment installed at customer sites, spare parts and field service visits need to be connected, because the customer's decision to renew depends on how these work together. We discuss this on our field service page, which covers managing technicians' visits to customer sites: https://www.gitogi.com/en/field-service/

We help Italian maintenance businesses digitise their operations and use AI through bespoke software, integrations, training and change management. The tools include AI agents: programs that carry out tasks by following instructions and using the information available. You can explore this further at https://www.gitogi.com/en/servizi/integrazione-ai/: the starting point is to understand where a business like yours loses time and information, so we can build a solution that fits the way you work.

The first step is practical: we start with a completed month's work, using that month's job reports, invoices and contracts. Together, we look at where service visits, contracts and invoices are disconnected. Then we measure the step that costs you the most. Before we build anything, we show you what it adds up to, so you can decide with the numbers in front of you.

If you recognise yourself in that owner wondering in October why a customer has left, write to us and tell us how you work: we'll start with a month of your service visits.

Frequently asked questions

Why can post-visit ratings be high even when some customers are unhappy?
Ratings only describe the customers who leave them, while some unhappy customers may leave without responding. In the study of more than 170,000 Samsung customer service chats, only 16.2% of customers left a rating, and those ratings were mostly positive. That does not tell you how satisfied the customers who left no rating were.
Is a customer who leaves no rating necessarily unhappy?
No. In the study of Samsung customer service chats, a program estimated that 54.5% of customers who had left no rating were dissatisfied. That is an estimate, rather than a rating those customers gave: silence alone cannot tell you whether a customer is happy.
What signals can you look at in a maintenance business besides ratings?
You can look at requests that never led to an appointment, repeat visits for the same fault, for example within thirty days, and contracts approaching expiry with no renewal offer. You can also compare renewal rates between customers who left a rating and those who did not, limiting the comparison to contracts that reached expiry in the same period and considering ratings left before expiry.
Can a survey also paint an overly negative picture?
Yes. As mTab explains, responses often come from very happy customers and very angry ones. A survey sent after a support request has been closed can therefore give negative experiences more weight than they have across your customer base as a whole.
Where does Gitogi start with a maintenance business?
We start with a completed month's work, using job reports, invoices and contracts to see where service visits, contracts and invoices are disconnected. Then we measure the step that costs you the most and, before we build anything, show you what it adds up to.

Sources

  1. Twilio, rapporto «Is AI Adding Value to the Customer Experience—or Taking It Away?»(accessed 11 November 2026)
  2. Twilio, blog(accessed 11 November 2026)
  3. Park, Cha, Rhim(accessed 11 November 2026)
  4. Bain & Company, «Breaking the Back of Customer Churn», Tom Springer, Charles Kim, Domenico Azzarello, Jeff Melton(accessed 11 November 2026)
  5. mTab, «Nonresponse bias», Mark Lummas(accessed 11 November 2026)

Updates on ERP, software and AI

Practical insights on ERP, bespoke software and AI in business workflows. No spam.

Redazione Gitogi

Gitogi Srl

LinkedIn

Let's talk about your next digital step.

A free 30-minute call to see where your business can grow — with Odoo, AI and custom software. No commitment.

Free, with no obligation. Francesco replies within 24 working hours. Be ready to explain how you work today and what's getting in the way.

Related articles